[Ffmpeg-devel-irc] ffmpeg.log.20181115

[burek]

[00:00:04 CET] <JEEB> and a basic FFmpeg, I'm not sure it will depend on anything else than system libs?
[00:00:41 CET] <AlexMax> Maybe.  Right now I just don't have any clue how to tell ffmpeg libraries to find the hardware decoder on my raspberry pi
[00:02:03 CET] <JEEB> can't say I found anything special in init code of the AVCodecContext in a player I poke around
[00:02:49 CET] <JEEB> and we can take a look at the init function in the decoder
[00:03:05 CET] <JEEB> AlexMax: also there's another function to go through decoders in a list
[00:03:18 CET] <JEEB> might want to make sure your library that you're linking against has the MMAL stuff
[00:03:36 CET] <AlexMax> I ran ldd on it, it look promising
[00:03:54 CET] <AlexMax> i saw some linking against some mmal libraries
[00:03:56 CET] <JEEB> http://git.videolan.org/?p=ffmpeg.git;a=blob;f=libavcodec/mmaldec.c;h=647a22ef7c07cc559de9f0d0ca2ad65d68cb7cf0;hb=HEAD#l349
[00:04:00 CET] <JEEB> this is the init function
[00:04:06 CET] <AlexMax> Ya know, I should probably do that
[00:04:25 CET] <AlexMax> I just didn't know if _I_ had to link against something extra to get mmal support or something
[00:04:48 CET] <JEEB> I'll check what the -codecs listing does in the mean time
[00:07:22 CET] <JEEB> ok, so it loops through AVCodecIDs in the binary
[00:08:43 CET] <JEEB> ok, lol. just uses a loop of av_codec_next() :P
[00:09:06 CET] <JEEB> checks if codec id matches what it wanted to loop though
[00:09:36 CET] <JEEB> and prints out those things
[00:10:27 CET] <JEEB> and in that av_codec_next loop you definitely do get an AVCodec pointer out
[00:10:35 CET] <JEEB> so if -codecs that print that bastard out
[00:10:50 CET] <JEEB> then you should be able to figure that thing out :P
[01:35:19 CET] <AlexMax> well this is distressing
[01:35:33 CET] <AlexMax> AVCodec* current_codec = NULL; current_codec = av_codec_next(current_codec);
[01:35:42 CET] <AlexMax> current_codec ends up null again
[01:35:47 CET] <AlexMax> no codecs?
[01:49:58 CET] <pink_mist> https://stackoverflow.com/a/50970973
[01:50:05 CET] <JEEB> AlexMax: if you are on older versions, have you tried registering them with av_register_all?
[01:50:43 CET] <JEEB> newer versions don't have the need, but older versions still had some dynamicness in there (which was only usable through private APIs, which is why it was removed)
[02:02:09 CET] <AlexMax> Aha, I think that worked
[02:02:55 CET] <AlexMax> Kind of strange that there is no trace of av_register_all in the original repository
[02:03:49 CET] <AlexMax> but this is ffmpeg 3.2
[02:10:00 CET] <AlexMax> that would explain why it didn't work for me on this linux box, but it did in my local MSYS2 setup
[03:11:19 CET] <ealdegue_> Hi ! anyone could help me with CMAF encoding and FFMPEG ? thanks !
[13:33:12 CET] <emacstheviking> I just screen recorded both monitors by accident. Is there a way I can split the video to reclaim just the right hand side as a separate file? I am reading the options now but there are so many!
[13:33:19 CET] <emacstheviking> worst case would be to convert all frames to jpeg, run them through imagemagick and stitch them back together.
[13:35:02 CET] <Mavrik> emacstheviking: "crop" video filter is made for that
[13:35:27 CET] <Mavrik> ffmpeg -i <yourfile> -codec:v libx264 -vf crop=<your crop parameters> -codec:a copy output.mp4
[13:35:34 CET] <Mavrik> look up the crop filter documentation for parameters :)
[14:37:11 CET] <CoreX> !filter
[14:38:32 CET] <CoreX> ohh he's gone if he comes back Mavrik you should give the person this https://ffmpeg.org/ffmpeg-filters.html#crop
[14:39:50 CET] <Mavrik> I was counting on the fact that he can google it :P
[14:47:47 CET] <durandal_1707> winny: 11
[14:47:58 CET] <durandal_1707> again, lol
[14:48:07 CET] <winny> durandal_1707: nice game
[15:24:35 CET] <ErwanD> Hey everyone
[15:24:44 CET] <ErwanD> I would like to use ffmpeg with opencl.  Working on windows 10 with mingw64.  I downloaded ffmpeg 4.1 sources, convert cuda opencl.lib to opencl.a. Then ./configure --arch=x86_64 --enable-opencl --prefix=/mydir Everything looks good. Then mingw32-make.exe, I had some errors related to ar fixed by modifying library.mak  Then install. But I don´t see any opencl.h header in mydir/include/libavutil
[15:24:49 CET] <ErwanD> Do you know if it´s normal ? How to ensure my build is "opencl ready" ?
[16:37:30 CET] <trashPanda_> Hello, does anyone have any experience using ffmpeg to restream a rtsp stream?
[16:38:07 CET] <trashPanda_> I am attempting to use the cli ffmpeg to re-stream a rtsp from a camera of mine, but I get a lot of "Error parsing AU headers" messages and no output stream
[16:38:29 CET] <trashPanda_> The rtsp stream does render correctly with ffplay however, is there something that could be messing with ffmpeg's ability to re-stream it?
[18:01:18 CET] <kepstin> ErwanD: there should be a few opencl filters included in your ffmpeg build if the opencl stuff was configured properly. (check by running ffmpeg -filters)
[18:19:49 CET] <Zexaron> Hey guys
[18:20:08 CET] <Zexaron> it's pretty clear WH is encoding videos different right https://www.politico.com/story/2018/11/14/white-house-trump-has-broad-discretion-to-regulate-journalists-access-989916
[18:20:52 CET] <Zexaron> if they're using CSPAN version, what is it that's doing that motion blur in particular, some kind of bad encoder setting with framerate, etc?
[18:21:20 CET] <Zexaron> the video is in the middle of the article
[18:22:04 CET] <pzy> I thought it was pretty clear the video was edited
[18:26:41 CET] <furq> didn't prison paul edit that and then show a clip of him editing the video in sony vegas to prove that it wasn't edited
[18:26:48 CET] <furq> he's not a smart man
[18:29:32 CET] <Zexaron> with all the codec artifacts it's hard to prove it
[18:30:32 CET] <Zexaron> any video could be doubted, any jpeg image, the common idea that "the blurring around the UFO proves it's fake", there's a lot of fakery on both sides, the UFO fakers and the fake assholes claiming everything's fake
[18:30:33 CET] <pzy> not really
[18:30:40 CET] <pzy> just an easy excuse to muddy the waters
[18:31:10 CET] <pzy> no one's mind has been changed and no discourse has been achieved, same ol' same ol'
[18:31:27 CET] <furq> there's a good rule of thumb you can use for situations like this
[18:31:37 CET] <furq> if two people each say the other one is lying, and one of them is donald trump, he's the one that is lying
[18:31:49 CET] <pzy> BUT BOTH SIDESSSSSSSS
[18:32:19 CET] <Zexaron> unfortunately theres absolutely no guidelines or laws on the governement or federal level on the technicalities sorrounding video evidence, what kind of quality and what kind of criteria it has to pass to be qualitifed for evidence in court (or I just don't know about it)
[18:32:39 CET] <pzy> that's up to 12 people to decide
[18:32:44 CET] <Zexaron> not both sides in that way, i was referring to various channels on youtube
[18:34:21 CET] <Zexaron> you shouldn't be able to use a crappy JPEG image to prove something in court, like a screenshot from a video which was recorded with a cellphone of the monitor which was playing back the video of a secruity camera
[18:34:44 CET] <Zexaron> only in russia tho
[18:34:59 CET] <pzy> that's why judges exist to determine permissibility
[18:35:10 CET] <furq> well presumably if this does go to court then they'll just request the actual footage from cspan
[18:35:44 CET] <Zexaron> none of them have ANY idea about the JPEG artificats, different techniques, photoshop, encoder differences, codecs, bitrate differences, color stuff
[18:36:29 CET] <pzy> that's why expert witnesses exist when necessary
[18:36:48 CET] <Zexaron> many of the things youtube kids jump to conclusions are infact artifiacts and glitches, the most common is the blur-border thing, it's a false positive
[18:37:35 CET] <Zexaron> yeah but it's not something usual, a lot of cort proceedings nothing happens automaticlaly, the defence has to ask for it
[18:37:53 CET] <Zexaron> demand it, etc, the court system is crap in so many ways
[18:38:12 CET] <ariyasu> the motion blue looks like iterpolation
[18:38:14 CET] <pzy> how many trials have you attended? our justice system is pretty amazing really
[18:38:23 CET] <Zexaron> the primary thing being, it takes money, which means, no money, no justice, that pretty much breaks it right there
[18:38:28 CET] <kepstin> regarding the sort of thing where digital media interacts with law, this is interesting reading: http://ansuz.sooke.bc.ca/entry/23
[18:38:30 CET] <ariyasu> nothing seems doctored
[18:39:01 CET] <trashPanda_> Its not doctored, just diff angle diff camera, decoded/encoded/compressed
[18:39:51 CET] <pzy> just throw more technical terms at people until they forget what the actual issue is
[18:42:43 CET] <kepstin> basically, the bits themselves can't tell you & don't record whether they've been edited or not, but the law cares, and so it's a tricky interaction :)
[18:42:56 CET] <Zexaron> yeah I was not sure if it's the same camera, but the point is pzy, there are SO MANY factors that it would take a team of experts, even if there's one obvious sign, you sill have a full analysis, just like NTSB, even if the people saw a missile laucnhed at the airplane, the cause of crash is obvious, but a full NTSB invesitgation with all the details would still be conducted
[18:43:32 CET] <Zexaron> and in that video analysis you would have to gather all the sources, raw copies, cameras themeselfs, all the technical details of the cameras, encoders, etc etc etc
[18:44:14 CET] <Zexaron> only then you start working looking for signs, and you need 10 year experience knowing all the false positives, that's what it takes to be impartial and fair at finding doctoring
[18:44:29 CET] <kepstin> in collecting all those details, you'll figure out the path that the video took to become the final product, which sometimes matters more than the actual contents of the video in question
[18:45:11 CET] <Zexaron> because the doctoring you see on youtube is easy and everyone can figure out, if it was really doctored they would to a lot better job than making it so easy that any random reader could notice
[18:47:38 CET] <Zexaron> And also, not underestimate the ignornace, especially young people, thousands of people still fall for easy youtube jokes, like the japan 2011 tsunami gas canister exploding being propagates as a "white ghost", and other stuff, it's very easy to fool people with genuine footage that's in low quality
[18:49:12 CET] <Zexaron> then you have more professional* fraud like Section51, which is a 17 year old stupid kid, makes totally fake youtube videos and gets millions of views sometimes
[18:49:35 CET] <pzy> is the video really the crux of this whole thing anyway? it has nothing to do with stated reasons and just serves to muddy the waters
[18:51:34 CET] <furq> this isn't really a new thing
[18:51:50 CET] <furq> remember all the psychics who convinced themselves that "auras" on photographs were ghosts and not just development errors
[18:52:13 CET] <furq> or all those moon landing guys who convinced themselves an eyelash that fell into a photo during processing was proof that a rock said "C" on it and was therefore proof that it was from a props department
[18:52:22 CET] <furq> those crazy props department and their clearly labelled props
[18:52:23 CET] <furq> s
[18:52:56 CET] <Zexaron> and then comes NASA with their mars rovers, maybe they got REAL RAW images but for public it's "RAW" JPEG only, if they wanted to prove they're on MARS etc etc, these JPEGs are so bad they can hardly be taken seriously.
[18:53:18 CET] <Zexaron> They would go to great lengths to doctor that footage, because of what's at stake and the consequences, but because they use JPEG and shit quality there's alot of possible false positives so this is an expert job:
[18:54:20 CET] <Zexaron> https://www.youtube.com/watch?v=Zj6uH-SqOZ8
[18:58:45 CET] <Zexaron> yeah, all these S marks on the Nasa Curiousy JPEGs, could it be another JPEG codec false-positive
[18:59:31 CET] <pzy> I feel like we did this 20 years ago already
[19:57:38 CET] <KombuchaKip> Zexaron: You might find this work I did on NASA Mars photo recovery technology. https://goo.gl/xB4CUi You can find the source here: https://github.com/cartesiantheatre/vlr
[19:57:46 CET] <KombuchaKip> Hey everyone. I'd really appreciate a code review of my custom AVIOContext for reading media from memory: https://pastebin.com/NGALFw8Y
[20:00:13 CET] <Zexaron> interesting
[20:14:30 CET] <KombuchaKip> Zexaron: Thanks. Unfortunately I don't have any DVDs left anymore for sale.
[20:21:40 CET] <pzy> really cool KombuchaKip
[20:22:16 CET] <pzy> when's the blu-ray version coming out? ;)
[20:22:49 CET] <KombuchaKip> pzy: Thank you guys. Hah, if you'd like to put one together I'm happy to put it on the parent project's website: https://www.avaneya.com
[20:23:11 CET] <KombuchaKip> pzy: It's a GPL game long away from completion I've been working on for years.
[20:27:51 CET] <pzy> is the processed but uncompressed video available somewhere, or would you have to run the utility yourself?
[20:36:27 CET] <trashPanda_> Does anyone have experience with RTSP streams?  I'm having an issue with restreaming a camera feed via ffmpeg.  I can display the feed just fine with ffplay.
[20:37:14 CET] <trashPanda_> When using ffmpeg, I get numerous "Error parsing AU Headers" messages until a timeout is reached
[22:45:56 CET] <BlackBishop> hi again awesome peeps.
[22:46:27 CET] <BlackBishop> any chance I can easily add a dts track to an mkv from an existing truehd/ac3 track ?
[22:47:07 CET] <JEEB> you can but not sure if you want. the only real reason to add a DTS track is to have bit streaming, but you already have two formats that are bit streamable
[22:47:51 CET] <BlackBishop> JEEB: the problem is that for some reason kodi on my tv has an issue with those but works just fine with dts !
[22:48:09 CET] <BlackBishop> no idea why !
[22:48:32 CET] <BlackBishop> and I'm trying to confirm the audio stream is the issue
[22:48:53 CET] <BlackBishop> I got 2 mkv files, identic video codec stream .. the only difference are the audio streams
[22:49:21 CET] <BlackBishop> 1 works ok (stutters from time to time a bit but it's ok) .. the other one starts to stutter and looses audio after a little while
[22:49:33 CET] <BlackBishop> it works ok when I pause/play or change audio tracks
[22:49:41 CET] <BlackBishop> anyhow, trying to confirm the issue
[22:50:25 CET] <JEEB> anyways, there's a lossy DTS encoder called "dca"
[22:51:21 CET] <JEEB> and it's marked experimental so you need -strict experimental after the input to use it
[22:56:21 CET] <BlackBishop> mkay..
[23:14:42 CET] <ILove47> Hi! I have no experience in video creation/editing at all and I'm not sure, it's the right channel to ask this, but I'll try :). If I should ask it in other channel - tell me where I can do that, please.
[23:14:42 CET] <ILove47> I want to programatically create a video with a scoreboard with row transitions and other pretty effects (like this: https://giphy.com/gifs/1Y6RMg4ptzpOAdlZhw/fullscreen). But I don't know what tools I'm able to use at all. I don't want to draw every single frame of that row transitions on my own but I don't know whether there are some tools or frameworks that can allow me doing this kind of things.
[23:16:01 CET] <ILove47> And of course, I forgot the question. :) What can I read and look for, so that will help me in reaching my goal?
[23:24:41 CET] <KombuchaKip> pzy: This is just using stills. The Viking mission only took stills.
[00:00:00 CET] --- Fri Nov 16 2018